3.7 Solar Cell Performance and Long-Term Stability I: Materials Library
3.7.1 Abstract
We present a newly developed setup that allows the combinatorial study of organic multi-layer devices for durability and degradation mechanisms. Arrays of 8×8 thin-film organic devices are prepared by vacuum deposition and investigated under vacuum conditions.
Measurements of 9 types of organic solar cells are presented. Continuous solar-like illu-mination at 80 mW/cm2 and a temperature of 80◦C provide realistic testing conditions.
The degradation of solar cells was investigated as a function of layer combination and layer thickness over a period of 800 hours. Different timescales of degradation were identified and attributed to the layer structure of the devices4.
3.7.2 Introduction
Since the first report of organic electro-optical devices by Tang et al. [Tan86, Tan87]
many efforts have been made to optimise the device parameters (film thicknesses, layer materials, etc.). The wealth of possible parameter combinations makes combinatorial methods a necessary tool for device development. Combinatorial device preparation has already been shown to be a very efficient tool [Sch99a,Sch99b], thus creating the demand for an equivalent testing procedure. Especially issues of degradation and lifetime can only be reasonably addressed with a combinatorial measurement setup. This may be one reason why only very few studies of the long-term behaviour have been carried out yet [Pet01, Hin01].
In this section we report the development of a combinatorial setup designed to study the degradation process of various organic thin-film devices in parallel. We present the simultaneous measurement of the electro-optical behaviour of 64 solar cells of nine different types over a period of more than 800 hours. We are able to distinguish two degradation mechanisms with different time scales.
3.7.3 Experimental
A two-dimensional array of 64 organic multi-layer solar cells was prepared by an evap-oration process starting with a 35 nm thick copper phthalocyanine (CuPc) layer as hole conductor onto an Indium Tin Oxide (ITO) coated glass substrate (76×76 mm). The ITO (Merck, 80 nm) was treated with an oxygen plasma before evaporation. On top of the CuPc layer, 25 nm of different perylene dyes (dimethyl perylene tetracarboxydiimide (DMPTI: 2,9-Dimethyl-anthra[2,1,9-def:6,5,10-d’e’f’]diisoquinoline-1,3,8,10-tetraone) and bisbenzimidazole perylene (BBIP: Bisbenzimidazo[2,1-a:1’,2’-b’]anthra[2,1,9-def:6,5,10-d’e’f’]diisoquinoline-6,11-dione (mixture with cis-isomer))) were evaporated as sensitisers in two different rows; one row was left empty. As a third functional layer, three columns
4The results of this chapter are published in [H¨an02]
70 Chapter 3. Combinatorial Techniques
Figure 3.34: Experimental setup: (a) Perspective view of the layered structure with elec-trodes. Illumination by a four lamp array and the electrical circuit of a single device are plotted schematically. (b) Top view of the 3×3 organic solar cell library and the 64 aluminium electrodes.
of 0, 15 and 30 nm of TiO2 were evaporated on top. This was done by heating Ti3O5 at an oxygen partial pressure of 2.5×10−5mbar, the subliming TiO reacting with the oxygen and forming a predominantly anatase crystal phase on the substrate. As the top layer an array of 8×8 aluminium electrodes was evaporated; this results in 9 device types with at least four devices per type (see Figure3.34), each 0.12 cm2 in size.
For simultaneous investigation of the 64 devices, the substrate was placed in a vacuum chamber at 5×10−5mbar. Electrical contacts to the electrodes as well as to the ITO were provided by gold-coated spring contacts. Electrical shorts were prevented by etching away the ITO in the contact area before evaporation. Four light bulbs, type OSRAM Ultra− Vitaluxr (300 W), provided homogeneous solar-like illumination with an irradiation in-tensity of 80 mW/cm2. (On a quasi-white background, there are emission peaks at 363, 404, 435, 545, 576, 624 and 728 nm).
The glass window of the vacuum chamber has a cutoff wavelength of 320 nm; no further optical filtering was done. The temperature inside the vacuum chamber reached about 80◦C. This corresponds to conditions found for inorganic solar cells in power generating setups. Every three hours, all devices were characterised by sequentially taking I–V curves from -0.2 to 0.6 V in steps of 0.02 V per second with aKeithley SMU 236. Switching between the devices was done by a home-built relay-multiplexer. During idle time, half the devices were short-circuited, while the other half were left open-circuited. The experiment was stopped after about 800 hours. During the measurement, 8 devices exhibited a short from the beginning on or showed strange behaviour in theI–V characteristic. They were excluded from the data evaluation and their values were replaced by the mean values of the corresponding group. Except for Figure3.35a, which nicely shows the homogeneity of the devices in a single group, all data presented is averaged over identical devices.
3.7. Solar-Cell Performance and Long-Term Stability I 71
Figure 3.35: (a) Current density ISC of all 64 devices after 4 and 800 hours of illumina-tion. The homogeneity of devices of the same type is clearly visible. The sensitising layer enhances ISC by one order of magnitude. The additional TiO2 layer does not increase the long-term device performance. (b) Time behaviour of ISC for the 9 device types. The data within respective groups has been averaged. Together with the data a single and a double exponential fit is shown. The decay times (in hours) are displayed in the bottom left-hand and the top right-hand corners respectively. Values of the less appropriate model are shaded grey.
3.7.4 Results and Discussion
Figure3.35a compares the short-circuit current (ISC) of all devices determined within the first measuring cycle and after 800 hours of illumination. For experimental reasons, the first measurement cycle was taken after 4 hours of illumination. We can see clearly that the initial performance of the solar cells is significantly improved by the sensitiser layers.
The effect of the additional TiO2 layer as an electron injection layer is less pronounced and different in sign for the two dyes. For the devices with DMPTI we find an increase in ISC which does not depend much on the layer thickness of TiO2. The contrary effect is observed for the BBIP layer, whereISC is slightly decreased. After 800 hours, the advantage of the TiO2 layer has nearly vanished in cells containing DMPTI. At the same time, the negative effect of TiO2 found in the BBIP devices has increased. Details regarding the initial effect of the TiO2 layer and on the evaporation technique are discussed in [The02]. The data presented in [The02] show an increase in efficiency even for BBIP, which at first glance seems to contradict our findings. We anticipate that the following discussion will identify a fast degradation process involving the TiO2 layer, so that the value of ISC taken after 4 hours of illumination will be smaller than the initial value taken independently on a different device without preceding thermal and optical stress.
We now turn to the discussion of the long-term stability of the multi-layer devices. Fig-ure3.35b shows the evolution of the short-circuit currentISC. We presentISC rather than the power conversion efficiency, as the former mainly reflects the exciton generation and charge separation probabilities whereas the latter additionally involves changes in the
72 Chapter 3. Combinatorial Techniques chemistry of the electrodes. We find a strong decrease in ISC for all nine device types, but the curves show a difference in shape. In order to analyse this decrease qualitatively, we fitted both a single and a double exponential to the sets of data. The statistical f-test with a 5% threshold was used to judge whether the more complicated model, i. e. the double exponential, is necessary to describe the data properly. Both fit curves are plotted together with the data. The decay times of the models in hours are given in the bottom left-hand and the top right-hand corners of the plots. We see that the devices without a sensitiser layer exhibit enhanced initial ISC values with increasing TiO2 thickness. After 800 hours, however, they show nearly the same ISC value. The decay times do not show a clear tendency with varying TiO2 thickness. The devices with a sensitiser layer but without a TiO2 layer exhibit a single exponential decay of three to four hundred hours. In the presence of a TiO2 layer we find an additional, shorter, decay time of around 30 hours.
The TiO2 layer thickness does not strongly influence the decay times. We note that for the double-exponential fits, fixing one decay time at the value of the device without TiO2 also results in consistent fits. The existence of more than one timescale is nevertheless evident.
The open-circuit voltage (VOC) is a second important parameter for the stability of the solar cells. Its behaviour with operation time is displayed in Figure3.36. It reflects changes in the HOMO and LUMO energy levels. Remarkably, we see two opposite tendencies at different timescales: Over a period of a few days, VOC increases for devices with no TiO2 layer. This period is shortened or vanishes completely in the presence of TiO2. All devices show a slight decrease in VOC on a timescale of several hundred hours. After 800 hours nearly all devices with TiO2 show a lower VOC than the corresponding devices without a TiO2 layer. Devices with both a TiO2 layer and a sensitiser layer show the strongest decay of VOC.
3.7.5 Conclusion
In summary, we find a long-term degradation process for all devices with a timescale of several hundred hours. Photo-bleaching may be anticipated as a possible reason, which would eventually lead to a lower exciton generation probability. Additionally, chemical changes of the electrode or injection layers may occur, which in turn lead to a decrease in electrical conductivity. Such changes are indicated by the long-term behaviour ofVOC and are most probably the result of the high temperature. The introduction of a TiO2 layer in thin-film organic solar cells can lead to initial enhancement of the performance. For CuPc/sensitizer cells the TiO2 layer introduces a new degradation mechanism and does not necessarily lead to an improvement in long-term performance. Our measurements suggest that the interface between the sensitiser and the TiO2 plays a key role in this process. Bearing in mind that the preparation of the TiO2 layer requires the reaction of the unstable TiO species with oxygen, this reaction may not have been complete and/or metastable crystals may have formed. Reordering or further oxidation of TiO may be involved. An oxidation of aluminium at the interface to TiO2 could also play a role.
A depth-profile analysis of the chemical composition could possibly help us achieve a deeper understanding of the degradation mechanism [Xin01] and is planned for future investigations.
3.7. Solar-Cell Performance and Long-Term Stability I 73
Open Circuit Voltage [ V ]
CuPc and DMPTI layer
0.4 CuPc and BBIP layer
TiO2 layer
0 nm 15 nm 30 nm
Figure 3.36: Open-circuit voltage vs. time: Devices with sensitiser and TiO2 layer show a decrease in VOC. Devices without this layer combination exhibit an initial increase and only a slight decrease for longer times if any. Together with the ISC data this leads to the assumption that the interface between TiO2 and the sensitiser layer is involved in the degradation mechanism.
We conclude that our setup is well suited to combinatorially characterise thin-film organic solar cells over a long period under realistic conditions. This is important for the under-standing of the degradation processes in thin-film devices and thus for optimising layer structure and layer thicknesses. Our results demonstrate that the initial performance, which is commonly reported for a new layer or material combination, is not sufficient to determine the suitability of a device. No differences were found between devices that were shorted and devices without current flow.
In future experiments we will study the influence of the illumination spectrum. A water filter has recently been built which is designed to cut out the infra-red spectrum and to allow temperature control. Further wavelength ranges can be cut out by the adding of dyes to the water. This should permit us to gain further insight into the degradation mechanism.
We finally note that the setup was originally developed for the long-term characterisation of organic light-emitting diodes. Electroluminescence can be detected via a CCD camera and each device can be driven at constant current. Preliminary results have been obtained
74 Chapter 3. Combinatorial Techniques for a single gradient device and work is in progress for combinatorial devices, as presented in [Sch99b, Bei02].